Papers by Manuela Braione
International Journal of Forecasting, 2020
In predicting conditional covariance matrices of financial portfolios, practitioners are
required... more In predicting conditional covariance matrices of financial portfolios, practitioners are
required to choose among several alternative options, facing a number of different
sources of uncertainty. A first source is related to the frequency at which prices are
observed, either daily or intradaily. Using prices sampled at higher frequency inevitably
poses additional sources of uncertainty related to the selection of the optimal intradaily
sampling frequency and to the construction of the best realized estimator. Likewise,
the choices of model structure and estimation method also have a critical role. In
order to alleviate the impact of these sources of uncertainty, we propose a forecast
combination strategy based on the Model Confidence Set [MCS] to adaptively identify
the set of most accurate predictors. The combined predictor is shown to achieve
superior performance with respect to the whole model universe plus three additional
competitors, independently of the MCS or portfolio settings.
Statistics & Probability Letters, Jul 20, 2016
We propose a scalar variation of the multivariate HEAVY model of Noureldin et al. (2012)featuring... more We propose a scalar variation of the multivariate HEAVY model of Noureldin et al. (2012)featuring a time-varying long run co-volatility component coupled with DCC dynamics.The new model outperforms the original HEAVY model by delivering more accurate multi-step-ahead predictions.
((Code to estimate the model downloadable in .zip format here: http://www.filedropper.com/tvlrmatlabpackage))
The Multiplicative Midas Realized DCC (MMReDCC) model of Bauwens et al (2014) generalizes the ReD... more The Multiplicative Midas Realized DCC (MMReDCC) model of Bauwens et al (2014) generalizes the ReDCC model of Bauwens et al (2012) by decomposing the conditional covariance matrix of returns into long-run secular and short-run transitory components in the spirit of Engle & Lee (1999). Given the multiplicative component structure of the model, estimation in realistic high dimensional settings can become infeasible due to an over-parametrization problem. In order to obtain a computationally feasible estimation procedure, we propose an algorithm that relies on the maximization of an iteratively re-computed moment-based profile likelihood function. The finite sample properties of the estimator are assessed via a comprehensive simulation study. Also, the results of two empirical applications are presented. The first demonstrates the accuracy of the algorithm in a low-dimensional setting while the second illustrates its effectiveness and practical usefulness in the implementation of high dimensional portfolio allocation strategies.
Novel model specifications that include a time-varying long-run component in the dynamics of real... more Novel model specifications that include a time-varying long-run component in the dynamics of realized covariance matrices are proposed. The modeling framework allows the secular component to enter the model either additively or as a multiplicative factor, and to be specified parametrically, using a MIDAS filter, or non-parametrically. Estimation is performed by maximizing a Wishart quasi-likelihood function. The one-step ahead forecasting performance is assessed by means of three approaches: model confidence sets, minimum variance portfolios and Value-at-Risk. The results illustrate that the proposed models outperform benchmarks incorporating a constant long-run component both in and out-of-sample. (Luc Bauwens), [email protected] (Manuela Braione), [email protected] (Giuseppe Storti) 1 Luc Bauwens and Manuela Braione acknowledge support of the "Communauté française de Belgique" through contract "Projet d'Actions de Recherche Concertées 12/17-045", granted by the "Académie universitaire Louvain".
Financial asset returns are known to be conditionally heteroskedastic and generally non-normally ... more Financial asset returns are known to be conditionally heteroskedastic and generally non-normally distributed, fat-tailed and often skewed. These features must be taken into account to produce accurate forecasts of Value-at-Risk (VaR). We provide a comprehensive look at the problem by considering the impact that different distributional assumptions have on the accuracy of both univariate and multivariate GARCH models in out-of-sample VaR prediction. The set of analyzed distributions comprises the normal, Student, Multivariate Exponential Power and their corresponding skewed counterparts. The accuracy of the VaR forecasts is assessed by implementing standard statistical backtesting procedures used to rank the different specifications. The results show the importance of allowing for heavy-tails and skewness in the distributional assumption with the skew-Student outperforming the others across all tests and confidence levels. Econometrics 2016, 4, 3 2 of 27
I briefly survey the recent literature on the high frequency data. I introduce the theory for rea... more I briefly survey the recent literature on the high frequency data. I introduce the theory for realized variation and covariation estimation and some specic issues related to the analysis of high frequency data, delving into the discussion of their strengths and weaknesses, the latter mostly due to the so called 'microstructure noise'. The attention is focused on the analysis of the most popular univariate volatility estimators and of their multivariate counterparts. In particular, I introduce the 'realized covariance' estimator followed by several bias-corrected estimators in the presence of noise. 2 Manuela Braione acknowledges support of the "Communauté francaise de Belgique" through contract "Projet d'Actions de Recherche Concertées" 12/17-045", granted by the "Académie universitaire Louvain". asset returns are negatively cross-correlated, in the sense that we expect that a past negative return will lead to a positive variation of volatility higher than the variation that would have been generated by a positive return of the same magnitude.
Matlab codes - Utilities by Manuela Braione
This code transform a (comma separated) csv file of data into an array of symmetric matrices, wit... more This code transform a (comma separated) csv file of data into an array of symmetric matrices, with a number J of series (specified in the vector 'excluded') not included in the final (n-J x n-J x T) matrix.
Transforms a [n(n+1)/2] dimensional vector (containing the lower triangular part of a matrix) int... more Transforms a [n(n+1)/2] dimensional vector (containing the lower triangular part of a matrix) into a symmetric [nxn] matrix.
function[RC_mat] = csv2array(T,N,r,c) %==========================================================... more function[RC_mat] = csv2array(T,N,r,c) %========================================================================== % This code transform a (comma separated) csv file into an array of % symmetric matrices. % The csv file must be organized in the following way: days per column and % on each row the matrix of day t in a vech form. % IMPORTANT: % Before running the code mind to SUBSTITUTE 'filename.csv' with the name % of the file to be opened which has to be placed in the same folder. %========================================================================== % INPUT % filename -the .csv file to be opened % T -horizon in days % N -number of assets % r -row from which starts the reading of the csv file % c -column from which starts the reading of the csv file %========================================================================== % OUTPUT % RC_mat -[N,N,T] array of symmetric matrices %========================================================================== % % Author: Manuela Braione, CORE-UCL 2014 % %========================================================================== vech = csvread('filename.csv',r,c); % Note that with a .xls file it's possible to use the following command:
function [results]=mcsEvaluate(mod,los1,varargin) %==============================================... more function [results]=mcsEvaluate(mod,los1,varargin) %========================================================================== % To be used to evaluate the results from the Model Confidence Set (Hansen, % Lunde, Nason 2011) when more than just one loss function is used. % Given a set of competing models indexed by numbers, this code compute % the absolute number of inclusions of each model in the MCS and the % corresponding overall percentage. % %
function [] = plotcov(h,C,S,G,varargin) %========================================================... more function [] = plotcov(h,C,S,G,varargin) %========================================================================== % This function plots in the same graph up to three different time % series not necessarily of the same length T. The mandatory input must be a % [nxnxT] array matrix. % h is the number of the asset to be plotted (h<=n).
Master Thesis Chapters by Manuela Braione
The study of volatility and covariation has become one of the most active and successful areas of... more The study of volatility and covariation has become one of the most active and successful areas of research in time series econometrics and economic forecasting in recent years. Thanks to the increasing availability of daily and intra-daily information on the returns of financial assets as well as computing power, different kinds of models have been proposed over the last two decades. In this thesis we deeply focus on two different classes of models, namely the Multivariate GARCH models, based on daily observed returns, and the Realized Covariance models, based on intra-daily data recorded at higher frequency. The reason behind this choice is due to the evidence that many researchers are currently interested in pointing out which class could be preferable in realistic financial settings and if using high-frequency information could really lead to significant estimation and forecasting improvements. We face this issue by analyzing a set of selected models and evaluating their profitability for portfolio optimization. The outline of the thesis is as follows. Chapter 1 can be split into two complementary parts. In the first one, we introduce the theory for realized variation and covariation estimation and some specific issues related to the analysis of high frequency data, delving into the discussion of their strengths and weaknesses, the latter mostly due to the so called 'microstructure noise'. We conclude the first part focusing on the analysis of the most popular univariate volatility estimators. The second half of the chapter is devoted to the discussion of their multivariate counterparts. In particular, we introduce the 'realized covariance' estimator and then we propose several bias-corrected estimators in the presence of noise. Chapter 2 is entirely devoted to the analysis of the Multivariate GARCH class of models. We review the existing literature and select the models to be developed in the application part. In particular, for each model, we present the parametric specification and illustrate a procedure for Quasi Maximum Likelihood estimation of the parameters of interest. Chapter 3 is inspired by the same logic of the previous chapter. In this chapter we introduce the dynamic models for time series of realized covariance matrices, namely the Realized Covariance models, which rely on the realized covariance estimator as a measure of co-volatility. Chapter 4 is the core of the thesis since it contains the original contribution of my work. It presents a detailed description of the steps through which the empirical analysis has been developed. First, the chapter illustrates the structure of our dataset containing 50 US stocks traded on the NYSE. Then, we illustrate the strategy used for the recursive estimation of model parameters and for the prediction of the conditional covariance matrices of asset returns. In particular we adopt a forecasting technique based on a rolling-fixed window scheme where the models are re-estimated at intervals of 100 days over a rolling window of fixed length set equal to 1500 days. After every model re-estimation, the predicted time series of conditional covariance matrices are used to compute alternative optimal portfolios for different values of target return. In the second part of the chapter we provide a brief review of the traditional instruments for measuring portfolio performance and discuss two different tests that can applied for testing the models forecasting accuracy: a simple regression based test and the usual Diebold-Mariano test. Then, in the last section, we discuss the empirical results of our analysis and offer some concluding remarks. The text is followed by an appendix collecting the tables and graphs containing results of the empirical application.
Competitions by Manuela Braione
Il gruppo tedesco dei "Cookies" si assicurano la vittoria nella competizione delle Casse di Rispa... more Il gruppo tedesco dei "Cookies" si assicurano la vittoria nella competizione delle Casse di Risparmio e Fondazioni europee. 36.642 squadre, di cui 1.091 italiane, hanno partecipato in tutta Europa all'edizione 2012 di Conoscere la Borsa.
Uploads
Papers by Manuela Braione
required to choose among several alternative options, facing a number of different
sources of uncertainty. A first source is related to the frequency at which prices are
observed, either daily or intradaily. Using prices sampled at higher frequency inevitably
poses additional sources of uncertainty related to the selection of the optimal intradaily
sampling frequency and to the construction of the best realized estimator. Likewise,
the choices of model structure and estimation method also have a critical role. In
order to alleviate the impact of these sources of uncertainty, we propose a forecast
combination strategy based on the Model Confidence Set [MCS] to adaptively identify
the set of most accurate predictors. The combined predictor is shown to achieve
superior performance with respect to the whole model universe plus three additional
competitors, independently of the MCS or portfolio settings.
((Code to estimate the model downloadable in .zip format here: http://www.filedropper.com/tvlrmatlabpackage))
Matlab codes - Utilities by Manuela Braione
Master Thesis Chapters by Manuela Braione
Competitions by Manuela Braione
required to choose among several alternative options, facing a number of different
sources of uncertainty. A first source is related to the frequency at which prices are
observed, either daily or intradaily. Using prices sampled at higher frequency inevitably
poses additional sources of uncertainty related to the selection of the optimal intradaily
sampling frequency and to the construction of the best realized estimator. Likewise,
the choices of model structure and estimation method also have a critical role. In
order to alleviate the impact of these sources of uncertainty, we propose a forecast
combination strategy based on the Model Confidence Set [MCS] to adaptively identify
the set of most accurate predictors. The combined predictor is shown to achieve
superior performance with respect to the whole model universe plus three additional
competitors, independently of the MCS or portfolio settings.
((Code to estimate the model downloadable in .zip format here: http://www.filedropper.com/tvlrmatlabpackage))